I think it's worth mentioning the differences between MAGRPO and MAAC's buffers. One is based on nodes, and one is based on the joint completions themselves.
Possible spaces where it should appear:
- Docs of MAAC, MAGRPO
- A statement of fair comparison in the docs (what things are aligned)
I think it's worth mentioning the differences between MAGRPO and MAAC's buffers. One is based on nodes, and one is based on the joint completions themselves.
Possible spaces where it should appear: